Confirmatory Adaptive Clinical Trial Design and Analysis
rpact supports the design, simulation, and analysis of
confirmatory clinical trials, from standard fixed-sample planning to
group sequential and adaptive methods. Sample size and power planning
and simulation are available for continuous, binary, survival
(time-to-event), and count data endpoints.
getDataset(est = ..., se = ...)We recommend three ways to learn how to use rpact:
- Use RPACT Cloud: demo.rpact.cloud
- Use the Vignettes: rpact.org/vignettes
- Book a training: rpact.com
RPACT Cloud provides guided, browser-based access to
selected rpact functionality, including study design,
sample size calculation, simulation, and reporting, without requiring
users to write R code: demo.rpact.cloud.
Practical examples and tutorials are available at rpact.org/vignettes. For function arguments, return values, and examples, see the function reference.
getDesignSet().getSampleSizeMeans(), or evaluate power for a specified
sample size with getPowerMeans().getSimulationMeans().getDataset() and
analyze them with getAnalysisResults().Plan a two-arm trial with a continuous endpoint, a one-sided significance level of 0.025, 90% power, and three analyses at 50%, 75%, and 100% of the planned information. Assume a mean difference of 0.5 and a common standard deviation of 1.
library(rpact)
design <- getDesignGroupSequential(
kMax = 3, informationRates = c(0.5, 0.75, 1),
alpha = 0.025, beta = 0.1, sided = 1, typeOfDesign = "OF"
)
sampleSize <- getSampleSizeMeans(
design = design, groups = 2, alternative = 0.5,
stDev = 1, allocationRatioPlanned = 1
)
summary(sampleSize)Sample size calculation for a continuous endpoint
Sequential analysis with a maximum of three looks (group sequential design), one-sided overall significance level 2.5%, power 90%. The results were calculated for a two-sample t-test, H0: mu(1) - mu(2) = 0, H1: effect = 0.5, standard deviation = 1.
| Stage | 1 | 2 | 3 |
|---|---|---|---|
| Planned information rate | 50% | 75% | 100% |
| Cumulative alpha spent | 0.0021 | 0.0105 | 0.0250 |
| Stage levels (one-sided) | 0.0021 | 0.0097 | 0.0215 |
| Efficacy boundary (z-value scale) | 2.863 | 2.337 | 2.024 |
| Efficacy boundary (t) | 0.631 | 0.415 | 0.309 |
| Cumulative power | 0.2928 | 0.6960 | 0.9000 |
| Number of subjects | 86.9 | 130.4 | 173.8 |
| Expected number of subjects | 130.9 | ||
| Stopping probability | 0.6960 | ||
| Efficacy stopping probability by stage (under H0) | 0.0021 | 0.0084 | |
| Efficacy stopping probability by stage | 0.2928 | 0.4033 |
Legend:
For a design without interim analyses, use
getDesignFixed() instead.
The following illustrative data represent two independent groups at each of three stages. Sample sizes, means, and standard deviations are stage-wise: each entry describes only the new observations collected in that stage. The sample sizes are chosen for illustration, independently of the planning result above.
trialData <- getDataset(
sampleSizes1 = c(40, 20, 20), sampleSizes2 = c(40, 20, 20),
means1 = c(0.2, 0.3, 0.5), means2 = c(0, 0, 0),
stDevs1 = c(1, 1, 1), stDevs2 = c(1, 1, 1)
)
analysisResults <- getAnalysisResults(
design = design, dataInput = trialData, directionUpper = TRUE
)
summary(analysisResults)Analysis results for a continuous endpoint
Sequential analysis with three looks (group sequential design), one-sided overall significance level 2.5%. The results were calculated using a two-sample t-test, equal variances option. H0: mu(1) - mu(2) = 0 against H1: mu(1) - mu(2) > 0.
| Stage | 1 | 2 | 3 |
|---|---|---|---|
| Planned information rate | 50% | 75% | 100% |
| Cumulative alpha spent | 0.0021 | 0.0105 | 0.0250 |
| Stage levels (one-sided) | 0.0021 | 0.0097 | 0.0215 |
| Efficacy boundary (z-value scale) | 2.863 | 2.337 | 2.024 |
| Cumulative effect size | 0.200 | 0.233 | 0.300 |
| Cumulative (pooled) standard deviation | 1.000 | 0.992 | 0.991 |
| Overall test statistic | 0.894 | 1.288 | 1.914 |
| Overall p-value | 0.1869 | 0.1001 | 0.0287 |
| Test action | continue | continue | accept |
| Conditional rejection probability | 0.0252 | 0.0337 | |
| 95% repeated confidence interval | [-0.459; 0.859] | [-0.196; 0.663] | [-0.020; 0.620] |
| Repeated p-value | 0.3515 | 0.1719 | 0.0338 |
| Final p-value | 0.0314 | ||
| Final confidence interval | [-0.016; 0.602] | ||
| Median unbiased estimate | 0.294 |
If the available summaries are cumulative, use the corresponding
cumulative prefixes, such as
cumulativeSampleSizes1, cumulativeMeans1, and
cumulativeStDevs1. For survival data, cumulative event
counts and log-rank statistics are commonly used; see
getDataset() for the required inputs. Analysis results and
the availability of confidence intervals depend on the chosen design and
endpoint; see getAnalysisResults() for details.
The terms survival and time-to-event refer to the
same endpoint type here. The corresponding function names use the suffix
Survival.
The most important rpact functions have intuitive names:
getDesign[Fixed/GroupSequential/InverseNormal/Fisher]()getDesignCharacteristics()getSampleSize[Means/Rates/Survival/Counts]()getPower[Means/Rates/Survival/Counts]()getSimulation[Means/Rates/Survival/Counts]()getSimulationMultiArm[Means/Rates/Survival]()getSimulationEnrichment[Means/Rates/Survival]()getDesignConditionalDunnett() for adaptive two-stage
multi-arm designsgetDataset()getAnalysisResults()getStageResults()RStudio / Positron / VS Code / Eclipse: auto code completion makes it easy to use these functions.
Result objects support familiar R functions such as
print(), summary(), plot(),
as.data.frame(), names(), and
length(). The available methods depend on the object class;
see the corresponding function reference.
Several utility functions are available, e.g.
getAccrualTime()getPiecewiseSurvivalTime()getNumberOfSubjects()getEventProbabilities()getPiecewiseExponentialDistribution()pi,
lambda and median, e.g., getLambdaByMedian()testPackage():
installation qualification on a client computer or company server (see
Installation
Qualification of rpact)For organizations using rpact in FDA/GxP-regulated
environments, RPACT offers support and formal validation documentation
for use on validated corporate computer systems. The documentation can
be customized and licensed for exclusive use by your company to support
applicable regulatory requirements.
The validation documentation also contains the personal access data
required to run the full test suite for installation qualification with
testPackage().
Contact RPACT to learn more.
RPACT Connect provides access to insights, downloads, and premium support for users and organizations working with RPACT software: connect.rpact.app.
rpact is free of charge and open source, licensed under LGPL-3. It implements a broad range of methods described in the monograph by Wassmer and Brannath (2025). Formal validation documentation is available to customers and supporting members through the RPACT service level agreement.
RPACT GmbH develops Statistical Tools for Drug Development. We develop and support R packages, Shiny applications, and cloud-based tools for clinical research, with services covering software development, validation, technical support, consulting, and training.
Our products include rpact and RPACT Cloud. We also maintain and support crmPack, the open-source package for dose escalation trial design, simulation, and analysis.
For more information, visit www.rpact.com.